Explainer: the challenges of grant assessment reform
There’s an argument that the most consequential development in UK research this year – in terms of how it will impact on researchers and research professionals – has been UKRI’s move to modernise grant assessment “for the age of AI”.
This wasn’t one policy change but rather a suite of measures, some of which have been long in gestation, others representing new ways of working. The context was UKRI’s strategy, which committed to a sharp reduction in grant processing times by 2031 – chief executive Ian Chapman told MPs earlier this year that UKRI as an organisation had “gulped” when he set the target.
The wider picture, as UKRI put it, was increasing competition and falling grant rates:
“We know that application volumes for some funding opportunities are very high, resulting in intense competition and low success rates.”
This has been partly ascribed to the ability of AI to facilitate the grant application process – but it also reflects tightening finances, with fewer opportunities available in some areas and greater pressure on researchers and institutions to obtain certain grants.
A programme of grant assessment reform didn’t come out of nowhere – the Tickell review of research bureaucracy called on funders to experiment with application processes to reduce burdens for applicants, and UKRI went on to commission a large-scale evaluation of approaches to peer review, which reported in December 2023.
However, the UKRI strategy and the subsequent announcements on modernisation have given this process renewed momentum – and it was notable that they were unveiled by new science minister Chris McDonald with some quite emotive language about funding competitions being a “value-destroying activity”.
UKRI’s news release stressed that it was committed to “maintaining the centrality of human judgement to funding decisions,” and that it was keeping an eye of equality considerations stemming from process changes. There wasn’t going to be a single silver bullet either:
“This is a fast-moving area, and the optimum system will require a portfolio of tools, not just one solution.”
The September announcement highlighted eight measures which will be adopted to both speed up decisions and reduce burden. This explainer will take a look at each of them – and consider some of the challenges they bring, both in mechanics and optics, which UKRI will have to overcome or make allowance for if it is to hit the ambitious targets in its strategy.
Two of the measures – empowering programme directors with more flexibility about assessment, and trialling “weighted multi-criteria” approaches to assessment – represent a combination of different things, and are unlikely to be deployed in one specific way (which is not to say that each does not merit further unpacking). This article concentrates on the remaining six discrete initiatives: more active demand management, larger and longer-term investments, in-person verification, distributed peer review, UKRI staff triage, and AI-assisted assessment.
Demand management
The first measure on the list is more active use of demand management, for “selected funding applications where volumes are expected to be high.”
As the Review of peer review framed it, demand management can be applied in different ways – at the level of the individual, this could involve limiting researchers to applying to fewer funding opportunities in total at a given funder (an approach which the review found to be very rare) or by way of constraining resubmissions. This latter approach is fairly common, but still can generate enormous controversy, as the European Research Council found earlier this year.
Typically, though, we’re talking about institutional limits – for example, UKRI’s recent proof-of-concept opportunity had limits by FTE staff submitted to REF, while the venture doctorates award allowed research organisations to submit only one application as lead.
Putting institutional limits on applications, and some of the pushback its extended use can garner, was covered on Research Agenda recently. The Review of peer review identifies two particular “hazards” with demand management: the shift of burden to institutions, in terms of what to take forward, and the potential for additional bias or less high-quality decision-making:
“The institution may opt for a more limited reviewing procedure, thus still reducing overall burden to some extent. However, there is also some anecdotal evidence that institutions may be less experienced in some aspects of selection processes, leading to suboptimal outcomes.”
UKRI has aimed to set out its “three-tier” demand management approach, which will see all funding opportunities make clear upfront whether demand management applies. However, it can be hard to predict in advance whether the scale of applications will warrant limits: UKRI’s recently opened policy fellowships opportunity notes that “if we get more applications than we can assess, we may start to limit numbers.”
Demand management is already used in many contexts – the challenge comes from scaling it or employing it more than the research community is anticipating. UKRI’s grant modernisation announcement states that it will make “more active” use of this measure, but it will do so by working in partnership with research organisations.
Larger and longer-term investments
The approach to bureaucracy reduction which Chris McDonald was keenest to highlight in his September speech to the Universities UK conference was the use of “fewer, longer awards.” UKRI describes this as
“focusing funding for strategic government and societal priorities (priority bucket two) on larger and longer-term investments to reduce the burden on applicants while supporting strategic outcomes.”
Larger grants will by definition reduce the number of funding opportunities which need to be reviewed and managed, but equally they increase the stakes for each given application.
And there is an open question about their applicability across the breadth of the research landscape, as British Academy chief executive Hetan Shah set out at our UKRI, transformed event earlier this month:
“Certainly in the humanities and social sciences, bigger isn’t always better, and you need a mixture of grants. At the British Academy we give £10,000 small research grants, and there’s huge appetite for those in our sector. That creates not just the seeds of new work that builds up to ESRC, AHRC or ERC grants, but complete projects in their own right, for that sum of money. I think government and UKRI can sometimes find it hard to think in that way.”
The indications are that this “larger and longer” approach is intended primarily for industrial strategy-related projects rather than discovery research. Here we might also note that the Budget promised “big bets” as part of the £1bn in UKRI funding which is being rerouted to clusters – this is likely, though not confirmed, to have a fair degree of overlap.
Adding devolution into the mix complicates things too. Larger grants make sense in mature areas where the government has clearly set strategic priorities – but do not seem such a natural fit for smaller, capacity-building work.
Interviews
Next up, “in-person verification”, with interviews being the clearest means of achieving this. UKRI suggests this will be “for higher-risk awards where additional assurances are required.”
Interviews as part of a grant assessment process are necessarily a resource-intensive step, and fall on the robustness side of the reform equation rather than helping to speed up processing times. As well as the time commitment that it supposes of panellists, interviews can also generate concerns about disadvantages for certain applicants. The Review of peer review highlighted this:
“Assessment through interviews can also be biased against certain personality types (for example, introverted, nervous, non-native speakers). In-person interviews may also pose difficulties for applicants with caring responsibilities or disabilities.”
As referenced above, UKRI has said that it will be “keeping a keen eye on the EDI implications of any changes” to its grant assessment processes. Interviews are already being used in funding opportunities, and scaling them further will conflict with commitments around efficiency too.
Distributed peer review
Distributed peer review has long been considered a way of tackling some of the challenges in the peer review process, not least the challenges faced both by funders and journals in attracting sufficient reviewers (especially in relatively niche areas). Nevertheless, it was still a surprise to see this particular approach get a shout-out in the 2025 Westminster Budget.
A good introduction to distributed peer review, which involves applicants reviewing each other’s submissions, can be found in this Wonkhe article and the Research on Research Institute guidance which it highlights. The metascience unit, a joint government-UKRI team, has been experimenting with and assessing the value of this approach for a while, and it has been employed in a number of UKRI-run funding calls in the last couple of years.
The metascience unit has suggested that it can give applicants more feedback, of a higher quality, and is time and resource efficient (for one thing, it ensures a reviewer response rate of close to 100 per cent, as refusal to review can result in auto-rejection). The unit’s same publication notes some of the potential drawbacks as well: a greater burden on applicants, for example, and less applicability in more heterogeneous funding calls.
It’s observed too that it can prevent reviews being conducted by those who were ineligible to apply but might have brought important expertise, such as those in industry or based elsewhere in the world. On a similar note, distributed peer review has also been criticised for presupposing a degree of uniform availability on researchers: those about to take parental leave, for example, struggle to fit in to the model.
Initial triage
Next up is the testing of initial triage by UKRI staff, another innovation which will be supported by the metascience unit. This could include funding council staff playing a “desk rejection” role in the way that editors of academic journals have long practiced.
The metascience unit’s A Year in Metascience report, referenced above, mentioned this back in summer 2025:
“We have heard anecdotal evidence that research council funding managers, who often have relevant academic expertise and credible judgement, can often tell when an application is likely to be rejected outright by a panel.”
Research council staff might, in fact, be better suited to this than reviewers in some circumstances, it is suggested:
“Many talented and dedicated UKRI staff are arguably better at identifying applications that are of a strategic fit to the call’s scope than reviewers, who, again anecdotally, can sometimes overindex on scientific quality or their own sense of important research questions, rather than strict adherence to the scope of a call.”
On the other hand, the metascience unit raises the possible controversies in allowing staff who are public servants to judge scientific quality – what it is seeking to obtain is robust evidence that UKRI programme managers would assess in a similar way to external peer reviewers. A “non-invasive shadow experiment” is being conducted to see whether this is the case, and the results will be interesting to see.
One other point to flag is that initial triage is tricky to square with UKRI’s goal of reducing the amount of its own resources which it spends processing grants. Desk rejection of applications reduces the amount of time spent by reviewers, and could speed up the overall timeframe as well – but it does require additional resource within UKRI.
AI
Finally on the list, there’s the elephant in the room: the potential use of artificial intelligence in assessments, “supported by appropriate controls and safeguards.” As the UKRI news item puts it, AI could be part of the solution as well as being part of the problem:
“Generative artificial intelligence… increasingly offers both opportunities and challenges for the grant application process. It might assist in both the writing and the reviewing of grant applications, but it also raises questions about privacy and the integrity of the review process.”
This is one area where the 2023 review of peer review has little by way of guidance to offer, saying that algorithms, AI or text recognition might be used to “aid allocation of reviewers to applications” – but that the technology has not been widely tested. Fast forward three years and there is a growing (though still hotly contested) recognition that AI could take on some of the review process itself.
We’ve already seen this autumn how the ill-judged use of AI in the assessment of applications can generate enormous blowback. This particular example was not a UKRI-run process – the funder is being much more cautious in determining what could work and how it would be presented to researchers, starting with a retrospective trial (again led by the metascience unit) to compare “what AI would have concluded from past applications against the decisions that were made.”
The metascience unit has recently said more about its ongoing work to come up a “basket” of AI-driven indicators which could speak to research quality – while at the same time stressing that whether this was deployed within UKRI would be a wider policy question beyond its remit.
Artificial intelligence use is undoubtedly the most controversial of the eight measures which URKI is scoping out for the purposes of reforming grant assessment. But as we have seen, none of them is without its drawbacks, and all have detractors. By committing in its strategy to bring down processing times quite so sharply, the funder has set itself a significant challenge – and a big part of this will be articulating its workings and rationale to the research community, and bringing researchers along with it.